Re: dynamic debug

[Johannes Berg <johannes@sipsolutions.net>]

On Fri, 2012-06-22 at 07:41 -0700, Joe Perches wrote:

> > > > Joe's conversion of mac80211 to pr_debug() was pretty much a disaster,
> > > 
> > > First I've heard of it.
> > 
> > Well I guess you didn't have the pleasure of having to work with bug
> > reporters whose mac80211 messages suddenly completely disappeared ...
> 
> Nor it seems the pleasure of interaction with a maintainer
> that forwards notices like this.

I told you before that I didn't like you wholesale converting it. I
might have been more open to it if you'd actually let me work on it and
convert some printk(KERN_DEBUG, ... to pr_info(). Now I'm stuck in a
situation where a lot of people will no longer see the messages they
expect, and I have to walk them through new, unexpected steps to make
them show up if I'm debugging a problem.

And heck, it's not even predictable since dynamic debug could be turned
on or off too.

> >  Doesn't really make a difference though. In reality,
> > most messages should be KERN_INFO anyway, so for those that aren't
> > hidden behind extra Kconfig options, we should use pr_info().
> 
> Why?  Aren't these then possibly some mixture of errors or
> notices or info level messages?

Yes, I believe that they are, but since you assumed that all KERN_DEBUG
messages were essentially useless by converting them to off-by-default,
I'm now changing it to assuming they're useful by converting them to
pr_info, we can change them back later.

However, most messages that aren't hidden behind Kconfig options are
actually useful, even if they were printed at KERN_DEBUG before.

johannes